Earticle

[목적] 본 연구는 축구 경기 이벤트 데이터를 기반으로 실시간 자동 중계문 생성 시스템을 구현하고, 템플릿 기반 방식과 KoAlpaca 기반 딥러닝 방식의 성능을 비교·분석하고자 하였다. [방법] StatsBomb Open Data에서 수집한 실제 축구 경기 데이터를 활용하였으며, 슈팅, 패스, 파울 등 주요 이벤트 1057건 중 약 120건을 정제하여 중계문 생성 시스템에 입력하였다. 템플릿 방식은 사전 정의된 규칙 기반 문장을, 딥러닝 방식은 KoAlpaca-Polyglot-12.8B 언어모델을 활용하여 중계문을 자동 생성하였고, 생성 결과는 BLEU 및 ROUGE 지표를 통해 정량적으로 평가하였다. [결과] 두 방식 모두 실시간 중계문 생성이 가능하였으나, KoAlpaca 기반 방식이 문장 표현의 다양성과 감정 전달 측면에서 더 높은 성능을 나타냈으며, BLEU-4 및 ROUGE-L 지표에서 템플릿 대비 향상된 결과를 보였다. [결론] 본 연구는 자동화된 스포츠 중계문 생성 시스템의 기초 구조를 제시하였으며, 향후 다양한 종목 및 언어 확장, 사용자 맞춤형 중계 서비스 개발로의 응용 가능성을 제시하였다.

[Purpose] This study aims to develop a real-time automatic sports commentary generation system using football match event data, comparing the performance of a template-based method and a deep learning-based KoAlpaca model. [Method] A total of approximately 120 key events, including shots, passes, and fouls, were extracted from StatsBomb Open Data for the 2018 FIFA World Cup match between Korea and Germany. Commentary was generated using rule-based sentence templates and the KoAlpaca- Polyglot-12.8B language model. The generated texts were evaluated quantitatively using BLEU and ROUGE metrics. [Result] Both approaches were able to generate commentary in real time. However, the KoAlpaca-based method demonstrated greater variation and contextual relevance in expression, achieving higher scores on BLEU-4 and ROUGE-L. [Conclusion] This study proposes a foundational framework for AI-based automatic sports commentary systems and suggests directions for future application, including expansion to various sports and languages and user-personalized commentary services.

공장에서의 대량 생산으로 사람이 직접 결함을 발견하는 것은 더욱 힘들어졌다. 결함 발견은 품질 관리 에서 필수적이므로 자동화된 결함 감지의 수요가 증가하고 있다. 이에 따라 컴퓨터 비전 기술을 이용하여 자동화 된 결함 발견으로 생산성을 향상하고 비용을 절감할 수 있는 연구가 관심을 받고 있다. 본 연구에서는 공기조화장치 내 부품을 생산하는 공장에서 캡과 밸브 위치에 대한 에러 검출을 하기 위해 템플릿 매칭과 딥러닝 모델 ResNet34 을 사용하였다. 실험 결과 딥러닝을 이용한 경우 템플릿 매칭보다 더 높은 정확도를 보였으며 top 이미지에서는 모든 불량 데이터를 검출할 수 있었고 side 이미지에서도 99% 이상의 불량 데이터 검출률을 기록하였다. 이러한 결과는 자동화된 결함 탐지 기술이 생산 관리의 효율적인 수단으로 활용될 수 있다는 가능성을 보여준다.

Mass production in factories has made it more difficult for humans to detect defects directly. As defect detection is essential in quality control, the demand for automated defect detection is increasing. Accordingly, research that can improve productivity and reduce costs through automated defect detection using computer vision technology is attracting attention. In this study, template matching and deep learning model ResNet34 were used to detect errors on the cap and valve position in a factory that produces parts in an air conditioning system. As a result of the experiment, when deep learning was used, it showed higher accuracy than template matching, and all defective data could be detected in the top image, and a detection rate of more than 99% was recorded in the side image. These results show the possibility that automated defect detection technology can be utilized as an efficient means of production management.

When it comes to current growth trend for the printing process in Asia printing market, offset printing, gravure printing and screen printing are reduced respectively –4%, –19%, –55%. In judging from the fact, the change in the printing production system from mass production on small amount to small production on mass amount is the biggest issue. For this reason, digital printing shows the significant growth. According to the increase of the growth 78% for electro photography way and 67% for ink-jet, it's not enough to catch up with digital printing which is increasing as time goes by to equip with hardware like as digital press. There's been necessary to install the BOD(Book on Demand) system which is the advanced and regular publication edit solution based on web-to-print model in the prepress, and it has made the BOD system be considered in relation to maximization of efficiency and production. Therefore, this research tries to step forward from the POD concept, which is refired to "Print along with the ordered quantities, the ordered appropriate time and the demanded place", so that it could expand the range of the printing/publishing environment using the BOD system, the order-made publication based on automatically operating template. And it tried to make the relation to digital web press on ink-jet method which is adequate to "mass production on small amount" with such advanced concept. This research also aims to use actively BOD solution model to promote the productivity of labor, and then to produce the printings across all related industries, which means to manufacture maximally the products on the shortest time at minimum place through PC equipments.

For improving automation calibration precision of multiple view texture collection in texture reconstruction, in this paper a new kind of calibration template using statistical distribution of surface texture is proposed. Firstly, there is a normalization processing be made according to statistical distribution of texture feature, such as species of color, quantity of geometric pattern and bump texture, etc. for the division of viewpoint. Then, the camera views of calibration template are assigned by using the normalization data with the calibration algorithm. Finally, the algorithm analysis is made by viewpoint division of four cases and the accuracy of our method can improve more than 10 percent. The experimental results show that, the proposed new calibration template method can better improve precision of 2D-3D registration and texture accuracy of reconstruction model than the uniform calibration template, while no obviously increasing the amount of calculation.

Due to Thai government policy give tablets to first grade students for use in learning and teaching. Mobile-augmented reality is used to help teach a powerful one. However, the most application programs are applied in English when used with the Thai alphabet is a decrease in performance. Therefore, we are developing template matching base-on ARToolKit used for Thai alphabet learning on mobile augmented reality. Especially, we are focusing on improvement character recognition for Thai alphabet learning, which enables extract character image as marker from mobile camera. To increasing the recognition rate of Thai alphabet base on ARToolKit by group similar shape for identifying each group of character. To dissimilarity, we propose a structured template matching technique for recognizing features such as knot of character. In this paper, we describe an abstract of our proposed method and present its experimental results.

오늘날 한국 사회에서 백내장 수술은 가장 빈번하게 이루어지는 수술중 하나이다. 백내장을 치료하기 위한 방법은 다양하게 발전해왔다. 이 중 오늘날 널리 쓰이는 방식은 인공 수정체를 이용하여 기존의 수정체와 교환하는 방법이다. 인공 수정체 삽입 시 미리 계산된 지점 및 각도로 정확하게 삽입하는 것이 중요할 수 있는 지점 표시가 필요 하다. 그러나 수술 중에 눈에서 분비되는 이물질 같은 요인에 의해 표시가 지워질 수 있다. 그러므로 이런 표시방법을 실시간으로 카메라 영상을 받음으로써 영상처리로 추적하는 방법이 필요하다. 본 논문에서는 템플릿 매칭과 버퍼를 이용한 특징점 추적 방법을 제안한다. 제안한 방법을 적용함으로 해당 수정체 삽입 위치를 정확히 추적하는 것을 확인하였다.

Today, a surgery for cataract of Korean society frequently come to operation. A method for treating cataracts have been developed in various ways. Widely used method is a method to use the artificial lens and replace it with the existing lens. The surgery can be inserted exactly according to the angle and points that are calculated in advance in the intraocular lens insertion is important. However, The lens insertion point can delete or blur the display due to such factors as foreign material coming out of the eye during surgery. Therefore, The lens insertion point needs a method of tracking the image processing by receiving the camera images in real time display method. In this paper, we propose a feature point tracking method by using template matching and buffer. a simulation results show that our ideas can track a feature point of the intraocular lens insertion.

Conducting polymers have been attracting considerable attention from both scientific and industrial perspectives by virtue of the beneficial electrical and optical properties originating from their unique ${\pi}-conjugated$ system. Many efforts have been devoted toward fabricating conducting polymer nanomaterials. Of the various synthetic methodologies, soft template method has emerged as a very promising tool in fabricating conducting polymer nanomaterials. For last a few years, our research group has intensively studied the fabrication and application of conducting polymer nanomaterials. In this talk, the recent achievement in the synthetic methodology based on the use of soft templates will be discussed.

Porous SiC samples were prepared with three types of wood (poplar, pine, big cone pine) by simply embedding the wood charcoal in a powder mixture of Si and $SiO_2$ at 1600 and $1700^{\circ}C$. The basic engineering properties such as density, porosity, pore size and distribution, and strength were characterized. The samples showed full conversion to mostly $\beta$-SiC with good retention of the cellular structure of the original wood. More rigid SiC struts were developed for $1700^{\circ}C$. They showed similar bulk density ($0.5{\sim}0.6\;g/cm^3$) and porosity (81~84%) irrespective of the type of wood. The poplar sample showed three pore sizes (1, 8, $60\;{\mu}m$) with a main size of $60\;{\mu}m$. The pine sample showed a single pore size ($20\;{\mu}m$). The big cone pine sample showed two pore sizes (10, $80\;{\mu}m$) with a main size of $10\;{\mu}m$. The bend strength was 2.5 MPa for poplar, 5.7 MPa for pine, 2.8 MPa for big cone pine, indicating higher strength with pine.

Electrochemical nitrogen reduction reaction (eNRR) offers a sustainable alternative for ammonia (NH₃) synthesis under ambient conditions, addressing the energy-intensive drawbacks of the traditional high-pressure, high-temperature Haber-Bosch process. Although eNRR offers a promising alternative for ammonia synthesis, significant challenges remain in achieving efficient performance. To address these, nanosized 2D layered transition metal phosphorus-sulfide materials, especially those incorporating cost-effective elements like Fe, are being investigated as potential electrocatalysts. In this study, FePS₃ was synthesized using a simple and environmentally friendly NaCl-template method. The eNRR performance of FePS₃ catalyst resulted in an NH₃ yield of 164.4 ㎍ h^-1 mg_cat^-1 with a Faradaic efficiency (FE) of 47% at an applied potential of -0.6 V vs. RHE under neutral pH condition (0.1 M Na₂SO₄). Notably, the electrolyte pH played a crucial role in influencing eNRR activity by modulating proton availability, which in turn affected both eNRR and the hydrogen evolution reaction (HER). This work demonstrates the successful synthesis of a low-cost, eco-friendly eNRR catalyst, emphasizing the potential of FePS₃ for efficient ammonia production.

Three cobalt-iron Prussian Blue Analogues (PBAs) nanotubes contained with different alkali metal cations of K, Rb or Cs, respectively, were prepared by using cetyltrimethylammonium bromide (CTAB)/ethanol-water micelles as soft templates. The products were characterized by energy-dispersive X-ray spectroscopy (EDS), X-ray diffraction (XRD), transmission electron microscopy (TEM) and scanning electron micrograph (SEM), which confirmed the composition of the substances and their unique nanotube structures. Furthermore, the formation mechanism of the PBAs nanotubes was discussed and provided useful insight for further synthesis of nanotubes of other Prussian blue analogues.

The paper was prepared spherical hexagonal boron nitride with different microstructures under flowing ammoniaatmosphere. The effect of boron source component ratios (molar ratio of borax and boric acid) on the phase composition ofthe sample were studied: the work also explored the effect of boron source component ratio on the microstructure andadsorption performance of h-BN. Spherical hexagonal boron nitride had corresponding to shell structure, mesoporousstructure, and solid structure when the ratio of boron source components respectively was 1 : 0.5, 1 : 1, and 1 : 2. Between 1 : 1and 1:0.5, the former had an lower adsorptions at low relative pressures. On the contrary, the adsorption of former wouldincreased at higher relative pressures and reached up 276.02 cm3 g?1, and the maximum adsorption of the latter reached only228.20 cm3 g?1. Under any relative pressure, the adsorption capacity was the lowest when the boron source composition ratiowas 1 : 2.

Transition metal oxide-based electrodes for lithium ion batteries have recently attracted much attention because of their high theoretical capacity. Here we report the electrochemical behavior of cobalt oxide nanorods as anodes, prepared by a template-free, one-step electrochemical deposition of cobalt nanorods, followed by an oxidation process. The as-deposited cobalt has a slightly convex columnar structure, and controlled thermal oxidation produces cobalt oxides of different Co/O ratios, while the original shape is largely preserved. As an anode in a rechargeable lithium battery, the Co/O ratio has a strong effect on initial capacity and cycling stability. In particular, the one-dimensional Co@Co_xO_y core shell structure obtained from a mild heat-treatment results in superior cycling stability.

GNSS has the disadvantage of being vulnerable to jamming, and thus, the necessity of jamming countermeasure techniques has gradually increased. Jamming countermeasure techniques can be divided into an anti-jamming technique and a jammer localization technique. Depending on the type of a jammer, applicable techniques and performance vary significantly. Using an appropriate jamming countermeasure technique, the effect of jamming on a GNSS receiver can be attenuated, and prompt action is enabled when estimating the location of a jammer. However, if an inappropriate jamming countermeasure technique is used, a GNSS receiver may not operate in the worst case. Therefore, jammer identification is a technique that is essential for proper action. In this study, a technique that identifies a jammer based on template matching was proposed. For template matching, analysis of a received jamming signal is required; and the signal analysis was performed using a spectral correlation function. Based on a simulation, it was shown that the proposed identification of jamming signals was possible at various JNR.

In this study we report a new method for the synthesis of a silica monolithic column bed with bimodal pores (throughpores and mesopores). The template induced synthesis method was used to direct bimodal pores simultaneously instead of the usual post base-treating method. Block polymer Pluronic F127 was chosen as a dual-function template to form hierarchically porous silica monolith with both macropores and mesopores. This is a simplification of the method of monolithic column preparation. Poly(ethylene glycol) was used as a partial substitute for F127 can effectively prevent shrinkage during the monolith aging process without losing much surface area (944 $m^2/g$ to 807 $m^2/g$). More importantly, the resultant material showed a much narrower mesopore size (centered at 6 nm) distribution than that made using only F127 as the template reagent, which helps the mass transfer process. The solvent washing method was used to remove the remaining organic template, and it was proved to be effective enough. The new synthesis method makes the fabrication of the silica monolithic column (especially capillary column) much easier. All the structure parameters indicate that monolith PFA05 prepared by the above method is a good material for separation, with the merits of much higher surface area than usual commercial HPLC silica particles, suitable mesopore volume, narrow mesopore size distribution, low shrinkage and it is easily prepared.

In object tracking, the template matching methods have been developed and frequently used. It is fast enough, but not robust to an object with the variation of size and shape. In order to overcome the limitation of the template matching method, this paper proposes a template update technique. After finding an object position using the correlation-based adaptive predictive search, the proposed method selects blocks which contain object's boundary. It estimates the motion of boundary using block matching, and then updates template. We applied it to IR image sequences including an approaching object. From the experimental results, the proposed method showed successful performance to track object.

Nitrogen-doped carbons have attracted much attention due to their novel application in relation to gas storage. In this study, nitrogen-doped porous carbons were synthesized using SBA-15 as a template, polypyrrole as the carbon and nitrogen precursor, and KOH as an activating agent. The effect of the activation temperature ($600-850^{\circ}C$) on the $CO_2$ adsorption capacity of the obtained porous carbons was studied. Characterization of the resulting carbons showed that they were micro-/meso-porous carbon materials with a well-developed pore structure that varied with the activation temperature. The highest surface area of $1488m^2g^{-1}$ was achieved at an activation temperature of $800^{\circ}C$ (AC-800). The nitrogen content of the activated carbon decreased from 4.74 to 1.39 wt% with an increase in the activation temperature from 600 to $850^{\circ}C$. This shows that nitrogen is oxidized and more easily removed than carbon during the activation process, which indicates that C-N bonds are more easily ruptured at higher temperatures. Furthermore, $CO_2$ adsorption isotherms showed that AC-800 exhibited the best $CO_2$ adsorption capacity of $110mg\;g^{-1}$ at 298 K and 1 bar.

압전 세라믹스는 전기에너지와 기계에너지를 상호 전환할 수 있는 성질을 기반으로 엑츄에이터 및 트렌스듀서 등에 사용되는 핵심 소재이다. 결정배향 성장법(TGG)은 다결정 세라믹스의 압전 특성을 획기적으로 향상시킬 수 있는 방법으로 많은 주목을 받고 있다. 하지만, TGG 방식을 통해 결정립을 배향하는 과정은 여러 단계의 최적화가 필요하기 때문에 입문자에게는 상당히 도전적인 기술이다. 따라서 이 기고문에서는 입문자의 입장에서 TGG를 가장 쉽게 접근할 수 있도록 납작한 모양의 템플릿을 합성하기 위한 용융염 합성기법 및 실제 결정배향을 위한 Tape Casting 과정 나아가 배향도를 향상시키기 위해 주의할 점에 이르기까지 TGG 전반에 대한 내용을 제공하고자 한다. 본 기고문이 TGG 방법에 대한 이해도를 높이고 활용 및 개선하고자 하는 모든 연구자들에게 정보 및 통찰을 제공할 수 있는 기본 참고서가 되기를 희망한다.

The templated grain growth (TGG) method has gained significant attention for its ability to produce highly textured piezoelectric ceramics with significantly enhanced performance, making it a promising method for transducer and actuator applications. However, the texturing process using the TGG method requires the optimization of multiple steps, which can be challenging for beginners in this field. Therefore, in this tutorial, we provide an overview of the TGG method mainly based on our previous published works, including its various processing steps such as synthesizing anisotropic-shaped templates with size and size distribution control using the molten salt synthesis technique, tape casting, and identifying key factors for proper alignment of the templates in the target matrix system. Our goal is to provide a resource that can serve as a basic reference for researchers and engineers looking to improve their understanding and utilization of the TGG method for producing textured piezoelectric ceramics.

Aluminum nanotube has been fabricated by a physical vapor deposition/atmospheric pressure injection using an anodic porous alumina film as a template. The pore external-, and inside diameters and the length of the aluminum nanotubes fabricated by this method are 60 nm, 35 nm and 2 $\mu$m, respectively. The structure of the fabricated aluminum nanotubes was examined by a kind of chemical treatment as extraction of copper on the cross-sectional area of these aluminum tubes in a mixed solution of $CuCl_2$ and HCl by difference of ionization tendency between aluminum and copper. The composition of the aluminum nanotube was identified by the two dimensional Hybrid Plasma Equipment Model (HPEM) employing the inductively coupled plasma.